LeCroy DDA 7 Zi series, Руководство оператора

Страница: 494 / 544

WavePro 7Zi
493
WP700Zi-OM-E-RevA
EPR4 2
E2PR4 2.31
However, the higher order PRML schemes need very complex circuits and decoders. While the Class IV partial
response (PR4) system works with three vertical levels of samples, extended partial response 4 (E2PR4) has
seven levels, and requires not only a higher resolution of ADC, but a complicated timing and gain recovery circuit
and sophisticated ML detector as well. Another disadvantage of the more complex PRML schemes is that they
are more sensitive to noise.
Principle of Equalization
The process of taking the more-or-less Lorentzian shaped head response to a magnetic transition and turning it
into a correctly shaped pulse is called equalization. This is of great importance, due to the need of the Viterbi
detector inside the PRML channel chip for correctly shaped pulses. Essentially, equalization is performed in the
read channel chip by a continuous time analog filter (CTAF).
Noise must be eliminated before sampling occurs, or else it becomes impossible to separate it from the head
signal. Since the head signal is typically noisy, and contains pulses that are not quite the desired shape, the DDA
provides an equalization filter to reduce much of the noise and reshape the pulses before it processes the
waveform. This filter is a digital implementation of a seven-pole, two-zero equiripple filter.
When you are using the filter, there are a number of parameters to be set. If the head signal has already been
acquired, the filter parameters can be set automatically by pressing the
Train Filter button. When this button is
pressed:
x
If the signal type is Peak Detect, the boost is set to zero and the -3 dB frequency is set to:
x
If the signal type is PRML (PR4, EPR4 or E2PR4) the -3 dB frequency is set to:
The best boost and -3 dB frequency are found by optimizing boost at the default -3 dB frequency, then optimizing
-3 dB at the better boost. Then optimize boost at the new -3 dB frequency, and optimize -3 dB frequency again at
the new boost. And then, if -3 dB frequency has changed by more than a small amount, optimize the boost one
final time. The goal for optimization is to maximize the mean of the 100 worst SAM values. A typical run will
recompute the filter, apply it, and run the Viterbi detector on the filtered waveform fifteen to twenty times. While
the filter training is in progress, a message is displayed showing the last boost and -3 dB settings and the mean of
the 100 worst SAM values at that setting.
The result of training is a close approximation to the best settings for our digital version of a CTAF on the current
acquisition, using the current setup of the FIR. The cleaner the waveform, the better the approximation will be.
The filter should be trained on a signal from a good read, those settings can be used for all reads in the same
zone.
Train Filter should be done with the acquisition stopped (press STOP on the DDA), so that the same waveform is
worked on each time; otherwise the search may be slow to converge. Train Filter can take a significant amount
of time, and it is recommended that it be done on relatively short waveforms of 50 or 100 kpoints. Once trained,
the memory length can be adjusted to the desired length. To ensure that the group delay is flat, the filter requires
at least five samples per bit cell. This is not a hard limit, but performance will degrade with fewer than five
samples per bit cell.
Alternatively, you can adjust the filter settings manually.
-3 dB Frequency
This is the actual -3 dB frequency of the filter. In most implementations of frequency cutoff (fc), the -3 dB point, if
Boost is 0 dB and group delay is 0%, is controlled by you. It is understood that the real -3 dB frequency will be
higher by some factor that depends on Boost and Group delay settings. Therefore changing Boost or Group delay
requires changing "fc" in order to keep the -3 dB point in the same place.